Mitochondrial DNA sequences of the Afro-Arabian spiny-tailed lizards (genus Uromastyx

Blackwell Science, LtdOxford, UKBIJBiological Journal of the Linnean Society0024-4066The Linnean Society of London, 2005? 2005 852 247260 Original Article Biological Journal of the Linnean Society, 2005, 85, 247 260. With 6 figures MOLECULAR PHYLOGENY OF UROMASTYX LIZARDS S. A. M. AMER and Y. KUMAZAWA Mitochondrial DNA sequences of the Afro-Arabian spiny-tailed lizards (genus Uromastyx; family Agamidae): phylogenetic analyses and evolution of gene arrangements SAYED A. M. AMER 1,2 * and YOSHINORI KUMAZAWA 1 1 Graduate School of Science, Nagoya University, Furo-cho, Nagoya 464 8602, Japan 2 Zoology Department, Cairo University, Giza, Egypt Received 4 December 2003; accepted for publication 16 August 2004 Approximately 1.7 kbp of mitochondrial DNA were sequenced from 29 individuals assignable to 11 Uromastyx species or subspecies and two other agamids. U. ocellata and U. ornata had an insertion between the glutamine and isoleucine trna genes, which could be folded into a stable stem-and-loop structure, and the insertion for U. ornata additionally retained a sequence similar to the glutamine trna gene. This corroborates the role of tandem duplication in reshaping mitochondrial gene arrangements, and supports the idea that the origin of light-strand replication could be relocated within mitochondrial genomes. Molecular phylogeny from different tree-building methods consistently placed African and Arabian taxa in mutually monophyletic groups, excluding U. hardwickii inhabiting India and Pakistan. Unlike previous studies based on morphology, U. macfadyeni did not cluster with morphologically similar Arabian taxa, suggesting convergent evolution to be responsible for the morphological similarities. Divergence times estimated among the Uromastyx taxa, together with geological and palaeontological evidence, suggest that the Uromastyx agamids originated from Central Asia during the Eocene and colonized Africa after its connection with Eurasia in the early Miocene. Their radiation may have been facilitated by repeated aridification of North Africa since the middle Miocene, and geological events such as the expansion of the Red Sea and the East African Rift Valley. 2005 The Linnean Society of London, Biological Journal of the Linnean Society, 2005, 85, 247 260. ADDITIONAL KEYWORDS: desert lizards gene organization molecular clock phylogeny trna. INTRODUCTION Spiny-tailed lizards or mastigures, of the genus Uromastyx, are herbivorous lizards highly adapted to desert environments, and grow to between 30 and 70 cm in total length. They are distributed throughout the arid regions of the Old World, ranging from northwest India to the Saharan region of Africa, including the Arabian Peninsula, as well as coastal Ethiopia and Somalia (reviewed in Wilms, 2001). Therefore, evolution of this lizard group may have been associated closely with palaeoenvironmental changes such as habitat fragmentation by orogenic or other plate tectonic activities, as well as local or global climatic *Corresponding author. E-mail: yasser92us@yahoo.com changes leading to aridification. Uromastyx is now registered in appendix II of CITES partly because of the danger of extinction by unlimited pet trading. Current classification of the genus Uromastyx is based mostly on external morphology, especially scalation (lepidosis) and body proportions (e.g. presence of intercalary scales between the annuli of the tail, tail-to-body length proportion, presence of femoral and preanal pores, and number of scales around the midbody) (Wilms, 2001). However, different opinions have been expressed about the intrageneric classification, including the placement of particular taxa either as full species, subspecies, or as a synonym of another taxon (Wilms, 2001 and refs. therein). In addition, researchers have been describing new species in recent years (Joger & Lambert, 1996; Mateo et al., 247

248 S. A. M. AMER and Y. KUMAZAWA 1998; Wilms & Böhme, 2000a, b). The most recent view (Wilms, 2001) lists as many as 16 species for this genus. In order to obtain independent data with which to assess the phylogenetic relationships within the genus Uromastyx, as well as to gain insight into palaeoenvironmental effects on the evolution of the desert lizards, it seems appropriate to employ a molecular approach. However, to the best of our knowledge, molecular phylogeny of Uromastyx has not been studied fully, except for a few studies in which several taxa were compared immunologically (Joger, 1986), or only a single representative species was sequenced to infer its relationship to the other agamids (Honda et al., 2000; Macey et al., 2000). We thus report here the first DNA molecular phylogeny that involves most African and Arabian taxa based on 1.7 kbp-long mitochondrial DNA (mtdna) sequences. We chose to target the mtdna region spanning from the trna Gln gene to the trna Tyr gene. This region has been sequenced for a number of other agamids (e.g. Macey et al., 1998, 2000; Schulte, Melville & Larson, 2003). The NADH dehydrogenase subunit 2 (ND2) gene in this region is one of the fastest evolving mitochondrial protein genes (see e.g. Kumazawa, Azuma & Nishida, 2004) and was expected to yield a number of base changes for the intrageneric phylogenetic analyses. MATERIAL AND METHODS SAMPLES A total of 27 Uromastyx individuals assignable to 11 taxa (see Appendix) were used in this study. Some of the samples were collected by one of the authors (S.A.) from the wild in Egypt (U. aegyptia aegyptia (Forskål), U. ornata Heyden and U. ocellata Lichtenstein) and in Sudan (U. dispar dispar Heyden and U. ocellata) and imported with the permission of CITES. In some cases, we preserved a frozen whole-body sample and deposited it in the National Science Museum, Tokyo (see Appendix for voucher numbers). However, because these taxa are easily identifiable in the field based on locality and morphological features (see Wilms, 2001), for the other samples we collected a small quantity of blood, after photographing the animals. Only extracted DNA was deposited in the museum for these samples. Samples for the remaining taxa were obtained through animal dealers or researchers without precise locality information. In most of these cases a wholebody specimen of a donated dead individual was deposited in the museum, while for a few specimens, only extracted DNA derived from blood taken from live animals was deposited. Various accurate information obtained through licensed animal importers on the origin of the samples (i.e. wild or bred), is given in the Appendix. Most samples originated from wild individuals free from artificial breeding with the other (sub)species. Although samples for U. acanthinura Bell (and potentially U. d. maliensis Joger & Lambert) originated from bred individuals, they were clearly identifiable based on morphological features (Wilms, 2001). Although we did not include several Uromastyx species from countries such as Iraq, Iran, Afghanistan and Somalia, these samples covered a broad geographical range of the genus, from India through to the Arabian Peninsula and to north-west Africa. The two outgroup taxa used for this study were Xenagama batillifera (Vaillant) and Calotes versicolor (Daudin). AMPLIFICATION AND DNA SEQUENCING Total genomic DNA was extracted from tissues according to Kocher et al. (1989) with minor modifications: DNA was extracted once by phenol and twice by chloroform/isoamyl alcohol and then ethanol-precipitated twice. Extraction from blood samples was conducted using a QIAamp DNA Blood Mini Kit (Qiagen). A negative control sample was carefully used through both the extraction and polymerase chain reaction (PCR) amplification procedures. Approximately 1.7 kbp of mtdna encoding the genes for trna Gln, trna Ile, trna Met, ND2, trna Trp, trna Ala, trna Asn, trna Cys and trna Tyr were amplified using the PCR primers shown in Figure 1 and Table 1. PCR reactions were set up in a total volume of 25 ml, containing 2.5 ml 10 Ztaq reaction buffer, 2 ml 2.5 mm dntps, 1.2 ml each of 10 mm primers, 0.25 ml (0.6 units) Ztaq DNA polymerase (Takara Shuzo Co.), and approximately 60 ng total genomic DNA. Thirtyfive PCR cycles were repeated, with denaturation at Figure 1. Position of primers used for amplification and/or sequencing. See Table 1 for the primer sequences; numbers of primers correspond to those in Table 1.

MOLECULAR PHYLOGENY OF UROMASTYX LIZARDS 249 Table 1. Primers used for PCR amplification and/or sequencing Name Sequence (5 to 3 ) Source PCR/sequencing primers 1: L4160m CGATTCCGATATGACCARCT Kumazawa & Nishida (1993) 2: L4184m TAATACACCTACTATGAAAAAAYTT Kumazawa & Nishida (1995) 3: H6313 CTCTYDTTTGGGGCTTTGAAGGC Sorenson et al. (1999) 4: H5934 CCCGACGCTGCAGGGTGCCAATGTCTTTRTGRTT Macey et al. (1997) Internal sequencing primers 5: rmet-1h GGCATGARCCCAAYTGCTT This study 6: rnd2-1l GCCCCMYTMCACTTCTGA Kumazawa & Endo (2004) 7: rnd2-2l TAGGRTGRACCATWTCAGCCAT This study 8: rnd2-3l CCTAATATACTACCTACGA This study 9: rnd2-1h ACTTCTGGWASTCAGAAGTG This study 10: rnd2-2h ATTGATGAGWAKGCTATRATTTTTCG Kumazawa & Endo (2004) 11: rnd2-3h TGTTCATATGGWTGTRGTGTC This study 12: rnd2-4h GTTATGGCTATRAGTRTTCCTA This study As a sequencing primer, H5934 was substituted by its shortened form H5942 (TGCAGGGTGCCAATGTCT). Primers designed specifically for Uromastyx. Primers designed for broader groups of reptiles. 98 C for 1 s, annealing at 50 C for 10 s (55 C for the outgroups and U. hardwickii Gray), and extension at 72 C for 30 s. Amplified fragments were purified with a High Pure PCR Product Purification Kit (Roche). They were directly sequenced with the PCR amplification primers and appropriate internal primers (see Fig. 1 and Table 1) using a DYEnamic ET Terminator Cycle Sequencing Kit (Amersham) on the Applied Biosystems 373 A DNA sequencer. The sequencing reactions consisted of 30 cycles of denaturation at 95 C for 20 s, annealing at 50 C for 15 s and extension at 60 C for 1 min. Complete nucleotide sequences of the genes were determined unambiguously by sequencing both strands. The nucleotide sequence data reported in this study will appear in the DDBJ/EMBL/GenBank nucleotide sequence databases with accession numbers AB113798 AB113826. PHYLOGENETIC TREE CONSTRUCTION Transfer RNA genes were identified and aligned by eye according to their standard secondary structure models (Kumazawa & Nishida, 1993). The alignment of the protein-coding gene ND2 was obtained with the aid of Clustal X (Thompson et al., 1997) executed at the amino acid sequence level, and finally corrected by eye. The alignment that included more outgroup taxa was deposited in the EMBL-Align database with accession number ALIGN_000633. Gap-containing sites, as well as ambiguous parts of the alignment, were subsequently removed so that 1503 alignable nucleotide sites remained for phylogenetic analysis. Phylogenetic analyses using nucleotide sequences were performed by three different methods: maximum-parsimony (MP) method using PAUP* 4.0b10 (Swofford, 2003) with heuristic searches by ten random additions of sequences and tree bisectionreconnection (TBR) branch swapping, neighbourjoining (NJ) (Saitou & Nei, 1987) using njboot in the Lintre package (Takezaki, Rzhetsky & Nei, 1995) with the Jukes-Cantor distance option, and maximumlikelihood (ML) using MOLPHY version 2.3 (Adachi & Hasegawa, 1996) with the HKY model and an MLestimated transition/transversion ratio parameter (3.48). In performing this ML analysis, we first obtained a tree by the star decomposition search, from which the local rearrangement search was conducted to provide the ML tree. Bootstrap probabilities for MP and NJ analyses were obtained from 500 replications, while those for ML analyses were local bootstrap values from 1000 replications using the RELL method (Adachi & Hasegawa, 1996). Statistical evaluation of alternative phylogenetic hypotheses by the Kishino Hasegawa test (Kishino & Hasegawa, 1989) was conducted using MOLPHY version 2.3 with the HKY model and the estimated transition/transversion ratio parameter (3.48). ESTIMATION OF DIVERGENCE TIMES Estimation of divergence times was carried out with either amino acid sequences of ND2 (339 sites), or nucleotide sequences of the trna Gln trna Tyr region (1503 sites). For the analyses using ND2 amino acid sequences, two distance options were used: gamma-

250 S. A. M. AMER and Y. KUMAZAWA corrected Poisson distances using tpcv in the Lintre package (Takezaki et al., 1995) and gamma-corrected ML distances using TREE-PUZZLE 5.0 (Schmidt et al., 2002) with the mtrev24 model and an MLestimated gamma parameter (alpha = 1.26). Such gamma-corrected distances for the mitochondrial protein sequences have been found to correct multiple substitutions at the same sites most efficiently to provide reasonable divergence times for divergences exceeding 100 million years ago (Mya) (e.g. Kumazawa, Yamaguchi & Nishida, 1999; Kumazawa & Nishida, 2000; Kumazawa et al., 2004). In order to use the nucleotide sequences for dating, 56 DNA substitution models were compared using Modeltest vers 3.06 (Posada & Crandall, 1998) to show that the TVM + I + G model best explained the DNA substitution process for the dataset. With this model and optimized parameter values, pairwise distances were calculated and used for the divergence time estimation. For both the amino acid and nucleotide analyses, reasonable calibration points proposed previously with the Laudakia agamids (Macey et al., 1998) were used. As the base substitution rate of mtdna sequences could vary from lineage to lineage, and the estimated rates could vary from method to method for the distance estimation, we did not apply the same clock value as in Macey et al. (1998) (0.65% per Myr) to our data. This is a major methodological difference from previous studies estimating divergence times (e.g. Schulte et al., 2003). We included sequences of 11 taxa of Laudakia (Macey et al., 1998) in our alignment to perform the relative rate test (the tpcv program) at all internal nodes of the best tree topology (see below and Macey et al., 1998 for the relationships within the Laudakia taxa) using Chamaeleo africanus as an outgroup. After excluding several taxa that may have had significantly (P < 0.01) accelerated or decelerated evolutionary rates (i.e. Calotes versicolor, Acanthosaura capra, Trapelus savignii and Leiolepis belliana), no rate heterogeneity was detected within and between the Uromastyx, Leiolepis and Laudakia species (data not shown). We then constructed the linearized tree using the tpcv program or averaged corresponding pairwise distances manually. These procedures enabled us to estimate reasonably the Uromastyx divergence times using the Laudakia calibration points and the clock rates estimated based on the gamma-corrected distances. We converted the distances to times by assuming that La. microlepis of Zagros diverged vicariantly from the northern taxa (two La. erythrogastra individuals and six La. caucasia individuals from different localities) by 9 Mya due to mountain uplifts caused by the Indian and Arabian collision (Macey et al., 1998). Use of other calibration points (e.g. the divergence of La. erythrogastra and La. caucasia 3.5 Mya due to the uplift of the eastern Kopet-Dagh mountains) gave us essentially the same results on the Uromastyx divergence times (data not shown). RESULTS MITOCHONDRIAL GENES OF THE UROMASTYX AGAMIDS We obtained complete nucleotide sequences for the mtdna region between the trna Gln and trna Tyr genes from 29 agamid individuals including 11 Uromastyx species or subspecies (see Appendix). When multiple individuals were available (for ten Uromastyx taxa except U. a. microlepis Blanford), two to four individuals were sequenced for the same taxon to examine polymorphic status. In almost all cases, sequences from the multiple individuals were identical or nearly identical with one or two base substitutions (data not shown) and we used only the first representative sequence for phylogenetic analyses (see Appendix to identify these sequences). Only for U. acanthinura and U. ocellata were two similar but clearly distinct sequence haplotypes found. The two haplotypes for U. ocellata were sampled from separate populations in Egypt and Sudan. These observations justify our phylogenetic study by showing good correspondence of molecular sequences for each taxonomic unit identified under the current classification (Wilms, 2001). Gene lengths were common for the agamids used in this study, except that the ND2 genes for U. a. aegyptia and U. a. microlepis had a 9 bp deletion at the 3 end (data not shown). No stop codon appeared within the ND2 coding region and there was no evidence to suggest that we collected nuclear or mitochondrial pseudogenes. The base composition was also similar among the agamid species although U. macfadyeni Parker (A = 34.4%, C = 31.7%, G = 14.2% and T = 19.8%) showed a base composition that was slightly different from the average (A = 35.1%, C = 29.6%, G = 13.1% and T = 22.2%) according to the 5% chi-square test (data not shown). The IQM trna gene cluster was rearranged to QIM for acrodont lizards (agamids and chameleonids) (Macey et al., 1997). The agamid taxa that we sequenced all had this QIM organization (Fig. 2B). While Leiolepis, like many other acrodont lizards, had a stable hairpin structure for the putative light-strand replication origin (O L ) between the trna Asn and trna Cys genes (Fig. 2B), none of the Uromastyx taxa examined did (Fig. 2C). Moreover, U. ocellata and U. ornata had an insertion of 128 and 59 bp, respectively, between trna Gln and trna Ile genes (Figs 2D, 2E and 3). Using PCR amplification with primers L4160m and rmet-1h and agarose gel electrophoresis, we confirmed that the insertions were present in all

MOLECULAR PHYLOGENY OF UROMASTYX LIZARDS 251 available individuals from U. ornata (N = 8) and U. ocellata (N = 7) but never in individuals from the other species (data not shown). There was 65% sequence identity between the insertions of U. ocellata and U. ornata (Fig. 3A), suggesting a common origin of this insertion. The remaining part of the insertion for U. ornata, showing weak sequence similarity (49%) to the original trna Gln gene (Fig. 3B), could be folded into a clover-leaf structure to produce another copy of the trna Gln gene or pseudogene (Fig. 4B). The inserted 128 bp sequences could take an alternative secondary structure with extremely stable and long stems (Fig. 4A). Figure 2. Evolution of mitochondrial gene organization in Uromastyx. A, typical vertebrate organization plesiomorphic to lizards. B, typical organization for acrodont lizards including Leiolepis and likely the direct common ancestor of Leiolepis and Uromastyx. C, typical Uromastyx organization in which the putative origin of light-strand replication (black box) disappeared from the WANCY trna gene cluster. D, organization for U. ornata and likely for the direct common ancestor of U. ornata and U. ocellata, which has an insertion containing a stem-and-loop structure (hatched box) and the second trna Gln gene or pseudogene (Q*). E, organization for U. ocellata in which Q* disappeared. See Figs 3 and 4 for sequences and secondary structures of the inserted region in U. ornata and U. ocellata. PHYLOGENY OF UROMASTYX Figure 5 shows an NJ tree obtained from the nucleotide sequences of the trna Gln trna Tyr region. The mtdna region produced a number of base substitutions (1076 variable sites and 948 parsimoniously informative sites out of 1503 alignable sites among the 25 sequences) and was effective for addressing the phylogenetic questions of Uromastyx. The three different analytical methods (NJ, ML and MP) provided identical tree topologies within Uromastyx and many clades were supported with strong bootstrap values. Taxon sampling for outgroup taxa may affect phylogenetic relationships and/or correct rooting of ingroup taxa. We therefore sampled as outgroups diverse agamid taxa in addition to a chameleon (Chamaeleo africanus). However, we found that the phylogenetic relationships within the genus Uromastyx were robust against removing or inserting several outgroup taxa (data not shown). In our molecular phylogeny, Leiolepis made a clade with Uromastyx, which is consistent with previous morphological (Moody, 1980; Böhme, 1982; Borsuk- Bialynicka & Moody, 1984; Frost & Etheridge, 1989) and immunological (Joger, 1991) studies. However, bootstrap probabilities to support this relationship were not very strong (57 88%). This result is therefore A B Figure 3. Nucleotide sequences of the inserted region between the trna Gln and trna Ile genes. A, alignment between the 128 bp insertion in Uromastyx ornata and the 59 bp insertion in U. ocellata (65% identity). B, alignment between the original trna Gln gene and its second copy within the inserted region for U. ornata (49% identity). Light-strand and heavystrand sequences are shown for A and B, respectively. Dots indicate identity with the first sequence and dashes denote a gap.

252 S. A. M. AMER and Y. KUMAZAWA Figure 4. Secondary structures of the inserted sequences found between trna Gln and trna Ile genes. The 128 bp insert for Uromastyx ornata can assume alternative secondary structures either with an extremely stable and long stem region (A) or with a clover-leaf structure for the second trna Gln gene (or pseudogene) and a stable stem-and-loop structure (B). The 59 bp inserted for U. ocellata may also assume a somewhat less stable stem-and-loop structure (C). Heavy-strand sequences are shown and numbers refer to the corresponding positions in their light-strand sequences shown in Fig. 3A. Bars in stems represent Watson Crick base pairs and dots stand for wobble G U pairs for RNA. not conclusive and is discussed later. Among the Uromastyx species, U. hardwickii diverged first, followed by divergence into two groups now inhabiting either North Africa or the Arabian Peninsula (including the north-eastern peripheral of Africa facing the Red Sea). For the purposes of this study we named these two clades African and Arabian groups. Our molecular phylogeny for the Uromastyx species (Fig. 5) is mostly consistent with morphology-based classification (Wilms, 2001). However, a few discrepancies can be seen: (1) U. macfadyeni clustered with North African taxa (U. acanthinura, U. geyri Müller, U. d. maliensis and U. d. dispar) and not with the Arabian group (U. ornata, U. ocellata and U. benti (Anderson)) as suggested morphologically; (2) U. aegyptia made a clade with the Arabian group (U. ocellata, U. ornata and U. benti) and did not cluster with the African taxa as suggested morphologically; (3) U. benti clustered with the U. ocellata U. ornata clade and not with U. ocellata as suggested morphologically. Estimated approximate divergence times are shown in Table 2. Leiolepis and Uromastyx were estimated to have diverged 40 50 Mya in the middle Eocene. U. hardwicki diverged from the other Uromastyx taxa during the Oligocene (25 29 Mya) and the divergence between the African and Arabian groups occurred in the middle Miocene (11 15 Mya). Within the Arabian group, U. benti was estimated to have diverged from the U. ocellata U. ornata clade 9 10 Mya and U. ocellata from U. ornata 7 8 Mya. Within the African group, U. macfadyeni diverged from the other African taxa 10 12 Mya, while the remaining African taxa diverged from each other more recently in the Pliocene or Pleistocene. U. geyri diverged from the U. acanthinura U. dispar clade 3 5 Mya, and U. acanthinura from U. dispar 2 4 Mya. DISCUSSION EVOLUTION OF THE MITOCHONDRIAL GENE ORGANIZATION IN UROMASTYX We found an insertion between trna Gln and trna Ile genes in U. ocellata and U. ornata. The inserted sequences of these species were similar to each other and thus may have a common origin (Fig. 3A). This section with the similarity could be folded into a stemand-loop structure in both species, albeit with less stability for the structure of U. ocellata (Fig. 4B, C). The 3 half portion of the U. ornata insert could be folded into a cloverleaf structure (Fig. 4B) and was similar in sequence to the original trna Gln gene preceding the insert (Fig. 3B). Therefore, it is likely that the 3 half portion of the U. ornata insert was created by tandem gene duplication. We cannot judge, without biochemi-

MOLECULAR PHYLOGENY OF UROMASTYX LIZARDS 253 Figure 5. Neighbour-joining tree constructed based on maximum likelihood distances from 1503 alignable nucleotide sites (the HKY model and transition/transversion ratio of 3.48). The tree was rooted with Chamaeleo africanus as an outgroup. Bootstrap probabilities are shown for neighbour joining, maximum likelihood and maximum parsimony analyses (from left to right). Underlined values mean that the branch was not reconstructed in the best tree topology by the corresponding analyses. Note that two distinct sequence haplotypes are included for Uromastyx acanthinura and U. ocellata. See Material and methods for more details on the analytical conditions. The nucleotide sequences taken from the database are: Chamaeleo africanus (accession No., AF448743), Chlamydosaurus kingii (AF128469), Physignathus lesueurii (AF128463), Acanthosoura capra (AF128498), Salea horsfieldii (AF128490), Trapelus savignii (AF128512), Leiolepis guentherpetersi (AF128461), Leiolepis belliana (U82689), Laudakia caucasia (AF028681) and Laudakia lehmanni (AF028677). cal experiments, whether the second copy is a functional trna Gln gene or a pseudogene. The secondary structure of the second copy nearly satisfied the criteria for functional mitochondrial trna genes (Kumazawa & Nishida, 1993). The only deviation from the criteria was that the nucleotide 3 next to the anticodon was T (Fig. 4B), not A or G as seen generally for the vertebrate mitochondrial trna genes (Kumazawa & Nishida, 1993). We suggest that, in the common ancestor of U. ocellata and U. ornata (approximately 7 8 Mya, see Table 2), a mtdna portion encompassing the trna Gln gene was tandemly duplicated and that the redundant second copy for the trna Gln gene was subsequently lost on a lineage leading to U. ocellata after divergence from U. ornata. However, the question arises as to why the inserted sequences homologous between U. ocellata and U. ornata (Fig. 3A) remain to be deleted in both lineages. An interesting possibility is that the remaining sequences, which can be folded into stem-and-loop structures (Fig. 4), retain some functional role, such as being a substitute for O L, which is apparently lost from the WANCY region. As two Leiolepis species retain the O L -like structure in the WANCY region and all Uromastyx species do not possess it, it is likely that the O L disappeared in the early Uromastyx lineage after its divergence from Leiolepis (Fig. 2C).

254 S. A. M. AMER and Y. KUMAZAWA Table 2. Divergence time estimates from molecular data ND2 amino acid seq. Nucleotide seq. Divergence at each node Pairs G-ML G-Poisson TVM + I + G Leiolepis vs. Uromastyx 12 45.3 39.3 ± 3.9 50.4 U. hardwickii vs. other Uromastyx 11 27.0 25.0 ± 2.6 29.2 African vs. Arabian groups 30 11.5 11.2 ± 1.3 14.5 U. aegyptia vs. other Arabian members 4 12.4 11.6 ± 1.5 13.6 U. a. aegytia vs. U. a microlepis 1 0.2 U. benti vs. U. ocellata-u. ornata 3 9.5 8.8 ± 1.3 10.4 U. ocellata vs. U. ornata 2 7.5 7.0 ± 1.2 7.7 U. ocellata (Egypt vs. Sudan) 1 0.5 0.5 ± 0.3 0.3 U. macfadyeni vs. other African members 5 10.9 10.1 ± 1.5 12.4 U. geyri vs. U. acanthinura-u. dispar 4 2.7 2.6 ± 0.6 5.4 U. acanthinura vs. U. dispar 4 2.1 2.0 ± 0.6 3.5 U. d. dispar vs. U. d. maliensis 1 0.6 0.5 ± 0.3 0.4 Divergence times based on gamma-corrected distances of the ND2 amino acid sequences and of the nucleotide sequences between trna Gln and trna Tyr genes were obtained as described in the Material and methods. Number of species pairs used for the time estimation. Means of divergence times among the corresponding species-pairs shown with one standard error. The ND2 amino acid sequences for U. a. aegyptia and U. a. microlepis are identical. As proposed by Macey et al. (1997), loss of the O L from the WANCY region may increase the chances for tandem duplication of trna genes, due to slippage in light-strand replication initiated from the secondary site in a trna gene. This raises the possibility of recreation of the O L in different regions of the mitochondrial genome. However, to the best of our knowledge, there has been no clear evidence to date for such recreation of the O L in different regions of mitochondrial genomes. It is difficult to know the structural requirements for O L function from only sequence comparisons (e.g. Wong & Clayton, 1985; but see also Macey et al., 1998). The question of whether the inserted stem-and-loop structures for U. ocellata and U. ornata (Fig. 4) actually function as a replicational origin should await independent biochemical characterization. CLASSIFICATION OF UROMASTYX We discuss the validity of the morphology-based classification of Uromastyx in the light of their molecular phylogeny. The molecular data support Uromastyx and Leiolepis forming a sister-group relationship as members of the most basal agamid lineage (57 88% bootstrap values; Fig. 5). Morphological studies since that conducted by Moody (1980) have pointed to this relationship, classifying them into a separate family Uromastycidae (Böhme, 1982), subfamily Uromastycinae (Borsuk-Bialynicka & Moody, 1984) or Leiolepidinae (Frost & Etheridge, 1989; Ananjeva, Dujsebayeva & Joger, 2001). To the best of our knowledge, molecular studies have not provided sufficient resolution on this matter. Partial 12S or 16S rrna gene sequences supported the sister relationship of Uromastyx and Leiolepis, but did not support the view that the clade represents one of the earliest agamid lineages (Honda et al., 2000). A tree constructed by Macey et al. (2000) using gene regions similar to ours did not even support the clustering of the two genera, although their taxonomic representations were somewhat biased towards agamids other than Uromastyx and Leiolepis. Therefore, there still seems to be uncertainty in the placement of Uromastyx and Leiolepis in the Agamidae, and this should be investigated further with increased molecular data. The molecular phylogeny within the Uromastyx species (Fig. 5) is in good agreement generally with the most recent view based on morphology (Wilms, 2001). U. hardwickii represents the most basal lineage of the genus and the remaining taxa investigated grouped into the African and Arabian clades. However, as outlined earlier, there were a few discrepancies with respect to the phylogenetic affiliation of U. macfadyeni, U. aegyptia and U. benti. Morphological studies (Moody, 1987; Wilms, 2001) pointed to certain similarities of U. macfadyeni to U. ocellata and U. ornata of the Arabian clade, with the last 12 21 tail whorls made up of a continuous scale row and with fewer than 260 scales around the midbody. In order to examine this further, we conducted the Kishino Hasegawa test to compare the log-likelihood values between the ML tree topology of Figure 5 and some competing hypotheses in which U. macfadyeni

MOLECULAR PHYLOGENY OF UROMASTYX LIZARDS 255 Table 3. Kishino-Hasegawa test for the phylogenetic position of Uromastyx macfadyeni Tree no. Sister group of U. macfadyeni lnl DlnL SE DlnL/SE 1 Other African taxa -19326.9 2 U. ocellata and U. ornata -19409.2 82.3 20.3 4.05* 3 U. ocellata -19440.7 113.8 24.1 4.72* 4 U. ornata -19446.1 119.2 23.6 5.05* The nucleotide sequences (1503 bp) for the 25 taxa were analysed as described in the Material and methods. User-defined unrooted tree topologies were as follows: tree 1 for the ML tree topology as shown in Fig. 5, (Outgroup, Hard, ((Mac, (Gey, ((Aca1, Aca2), (Dis, Mali)))), ((Aeg, Mic), (Ben, ((Oce1, Oce2), Orn))))); tree 2, (Outgroup, Hard, ((Gey, ((Aca1, Aca2), (Dis, Mali))), ((Aeg, Mic), (Ben, (Mac, ((Oce1, Oce2), Orn)))))); tree 3, (Outgroup, Hard, ((Gey, ((Aca1, Aca2), (Dis, Mali))), ((Aeg, Mic), (Ben, (Orn, ((Oce1, Oce2), Mac)))))); and tree 4, (Outgroup, Hard, ((Gey, ((Aca1, Aca2), (Dis, Mali))), ((Aeg, Mic), (Ben, ((Oce1, Oce2), (Mac, Orn)))))). Refer to Fig. 5 for the tree topology of the outgroup and see the legend of Fig. 6 for abbreviations of Uromastyx taxa. Oce1 and Oce2 mean U. ocellata from Egypt and Sudan, respectively. Natural logarithm of the likelihood value. Difference in lnl from that of the ML tree. A standard error in lnl. An asterisk means that the corresponding phylogenetic hypothesis can be statistically rejected at the 1% significance level by the standard criterion of DlnL/SE > 2.58. Figure 6. Hypothetical radiation schemes for Uromastyx and possibly relevant geological events. Approximate distribution range for each taxon (Wilms, 2001) is shown with its abbreviated name: Hard (Uromastyx hardwickii), Aca (U. acanthinura), Mali (U. d. maliensis), Gey (U. geyri), Dis (U. d. dispar), Oce (U. ocellata), Mac (U. macfadyeni), Aeg (U. a. aegyptia), Mic (U. a. microlepis), Orn (U. ornata) and Ben (U. benti). clusters with either or both U. ocellata and U. ornata (Table 3). The results clearly rejected the competing hypotheses at the 1% significance level. We therefore suppose that the apparent similarities between U. macfadyeni and the latter two species may be due to convergent evolution through sharing similar ecological or climatic environments. These species are now distributed in neighbouring areas surrounding the Red Sea and the Gulf of Aden (Fig. 6). U. aegyptia was considered to have a phylogenetic affinity to the African group because of their common features in external morphology, such as the last two to five tail whorls being made up of a continuous scale row (Moody, 1987; Wilms, 2001). However, this species has much higher midbody scale counts than do members of the African group (Wilms & Böhme, 2000a). The phylogenetic affiliation of U. aegyptia in the Arabian group (Fig. 5) may therefore be possible, though this is not conclusive due to the low bootstrap values (52 57%). It is noteworthy that the serological studies by Joger (1986) also suggested a closer relationship of U. aegyptia to U. ocellata and U. ornata than to

256 S. A. M. AMER and Y. KUMAZAWA U. acanthinura and U. geyri. Finally, the molecular phylogeny (Fig. 5) placed U. benti as sister to the U. ocellata U. ornata clade, while U. benti was morphologically regarded as a sister taxon of U. ocellata with the exclusion of U. ornata (Wilms, 2001). We favour the molecular view in this respect because of its strong bootstrap values (98 99%). Treatment of a taxon as either a species or a subspecies has been changed frequently for members of the genus Uromastyx, and this may still be controversial. For example, Mertens (1962) classified U. geyri and U. dispar as subspecies of U. acanthinura, while Moody (1980) recognized them as full species. Thomas Wilms followed the former view in his book (Wilms, 1995) but later changed to the latter (Wilms, 2001). He also once recognized U. ocellata, U. ornata and U. macfadyeni as subspecies of U. ocellata (Wilms, 1995), while all of them have recently been treated as independent species (Wilms & Böhme, 2000c; Wilms, 2001). Conversely, U. d. maliensis, once recognized as an independent species (Joger & Lambert, 1996), has been revised to be a subspecies of U. dispar (Wilms & Böhme, 2000b). In all the above-mentioned points, our molecular phylogeny (Fig. 5) supports the most recent view in Wilms (2001), not only for the phylogenetic relationship but also for the level of divergence. The divergence times between U. geyri, U. acanthinura and U. dispar, as well as between U. ocellata and U. ornata, were much larger compared with those between U. d. dispar and U. d. maliensis (Table 2). HISTORICAL BIOGEOGRAPHY Extant Uromastyx taxa of the African and Arabian groups are distributed allopatrically, with a considerable overlap of distribution only between U. geyri and U. d. maliensis (Fig. 6). In this section, we discuss how the Uromastyx species radiated and migrated to their current habitats, based on our molecular data together with geological and palaeoenvironmental evidence. The evolutionary framework thus constructed using representative sequences from each taxon will provide a basis for future phylogeographical analyses using many individuals with detailed locality information. Our molecular analyses suggested that Uromastyx and Leiolepis are sister genera (Fig. 5) and that they diverged from each other in the middle Eocene (40 50 Mya; Table 2). The oldest fossil record closely associated with these genera is Mimeosaurus of the Upper Cretaceous Eocene of Mongolia (Moody, 1980), and is consistent with the above-mentioned divergence time. Moreover, all the extant species included in the genus Leiolepis inhabit south-east Asia. U. hardwickii, having diverged from the most basal position of the Uromastyx phylogeny (Fig. 5), inhabits south Asia. Thus, it is likely that direct ancestors of these genera lived in central south Asia, from where the genus Uromastyx originated and migrated westward towards the hot and arid habitats suitable for their lifestyle (Fig. 6). From the late Cretaceous to the early Miocene (18 100 Mya) the African continent, including an area of the current Arabian Peninsula, had long been isolated from other continents (Rögl, 1998). Geological and palaeontological evidence (Rögl, 1998; Harzhauser, Piller & Steininger, 2002) consistently shows that plate tectonic activities connected Africa to Eurasia through closure of the Eastern Mediterranean seaway by approximately 18 Mya (the Gomphotherium Landbridge). This landbridge later became disconnected temporarily, but it has been continuously present since ~15 Mya. Our molecular analyses suggested that the Asian (U. hardwickii) and Afro-Arabian (the other species used) taxa diverged 25 29 Mya (Table 2). This is much earlier compared with the estimate for the formation of the Gomphotherium Landbridge. We therefore speculate that there was an initial stage of radiation of the Uromastyx lizards in the eastern Middle East before the formation of the landbridge, and that predecessors of the Afro-Arabian taxa derived from one of these diversified lineages. A few taxa from the eastern Middle East (e.g. U. loricata of Iraq and Iran and U. thomasi of Oman) were not included in our study. Morphological (Wilms, 2001 and refs. therein) and immunological (Joger, 1986) studies have suggested that these taxa diverged from basal positions next to that of U. hardwickii in the Uromastyx phylogeny. We estimated the African and Arabian groups of Uromastyx to have diverged 11 15 Mya (Table 2). North Africa was subjected to climatic changes towards aridity in the middle Miocene. Dry and open woodlands and the savanna emerged to interrupt the continuous African forest by the late middle Miocene (McClanahan & Young, 1996; MacDonald, 2003), and this may have facilitated faunal and floral exchanges eastwards and westwards (e.g. Fu, 1998; Caujapé- Castells & Jansen, 2003). Predecessors of the African Uromastyx group may have migrated to new xeric habitats in North Africa and diverged from the Arabian taxa (Fig. 6). The Arabian Peninsula began to separate from the remaining part of the African continent in the early Miocene as a tectonic consequence of the Afar mantle plume leading to the formation of the Red Sea, the Gulf of Aden and the East African Rift Valley (Girdler, 1991; Pudlo, Shandelmeier & Reynolds, 1997). This accompanied considerable uplifting of some mountain systems along the plate boundaries. The East African Rift Valley and associated mountain systems

MOLECULAR PHYLOGENY OF UROMASTYX LIZARDS 257 may have become terrestrial barriers, isolating U. macfadyeni of northern Somalia from its sister taxa of the African group. The estimated divergence time (10 12 Mya; Table 2) seems consistent with geological (Girdler, 1991) and palaeontological (Coppens, 1994) data in supporting this idea. Within the Arabian group, U. aegyptia first diverged from the other members 12 14 Mya (Table 2) and this species is now widely distributed in the Arabian Peninsula and northern Egypt. Molecular data also suggested that U. benti diverged from U. ocellata and U. ornata 9 10 Mya and that the latter two species diverged from each other 7 8 Mya (Table 2). A possible geological factor that may have been associated with the former divergence is the elevation of the Yemen Plateau (Geoffroy, Huchon & Khanbari, 1998). However, the considerable uplifting of the Yemen Plateau may have already occurred in the early Miocene, somewhat earlier than the estimated divergence time (9 10 Mya). We therefore withhold a conclusion on this matter until more molecular and geological data become available. Geological influence on the divergence between U. ocellata and U. ornata seems much clearer. Since the current distribution ranges for U. ocellata and U. ornata are separated by the Red Sea, their divergence has been hypothesized to be due to the habitat fragmentation caused by the expansion of the Red Sea (Wilms, 2001). Our study supports this hypothesis by showing that the estimated divergence time between the two species (7 8 Mya) corresponds well to the geological timing for the expansion of the Red Sea (Girdler, 1991; Pudlo et al., 1997; refs. therein). The oceanic accretion may have started in the middle Miocene (12 13 Mya), while evidence from sedimentary rocks suggests that seawater had come in to the northern region of the Red Sea by at least 5 Mya (Ross & Schlee, 1973). Our molecular data pointed to much more recent times for the divergences between members of the African group other than U. macfadyeni (Table 2). By the early Pliocene (around 5 Mya), the trend for a cooler and drier environment was well established in North Africa with expansion of the grassland (MacDonald, 2003). After 2.8 Mya, there were repeated global cycles of warming and cooling and this is believed to have accelerated speciation for a variety of terrestrial and marine animals (Agusti, Rook & Andrews, 1999; demenocal & Brown, 1999). Especially during the late Pliocene (around 2.4 and 1.8 Mya), further cooling and drying resulted in a major expansion of grassland and desert environments (McClanahan & Young, 1996). We suggest that such climatic fluctuations could have caused the habitat fragmentation and isolation of local populations, leading to the speciation between U. geyri, U. acanthinura and U. dispar. ACKNOWLEDGEMENTS We are grateful to Messrs. Kensuke Iwata, Kosho Yagi, Usama AlBendary, Kamal Galander, and Drs Tomohide Yoshimura, Samy El-Fellaly and Abdel Rahman Tawfik for their assistance in collecting samples. We also thank Messrs. Riosuke Aoki and Akihiro Koshikawa for critically reading the manuscript and providing comments on it. Gratitude is extended to Dr Yuji Orihashi for providing information on the geological settings of the Arabian Peninsula, and Dr Hideki Endo for his help in depositing our specimens in the National Science Museum, Tokyo. This work was supported in part by a postdoctoral fellowship from the Japan Society for the Promotion of Science to S.A., and by grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan (grants 14540641 and 14902202). REFERENCES Adachi J, Hasegawa M. 1996. MOLPHY version 2.3: programs for molecular phylogenetics based on maximum likelihood. Computer Science Monographs 28: 1 150. Agusti J, Rook L, Andrews P. 1999. Chapter 1: introduction. In: Agusti J, Rook L, Andrews P, eds. The evolution of Neogene terrestrial ecosystems in Europe. London: Cambridge University Press, 1 6. Ananjeva NB, Dujsebayeva TN, Joger U. 2001. Morphological study of the squamate integument: more evidence for the metataxon status of Leiolepidinae. Journal of Herpetology 35: 507 510. Böhme W. 1982. Über Schmetterlingsagamen, Leiolepis belliana belliana, der malayischen Halbinsel und ihre parthenogenetischen Linien (Sauria, Uromastycidae). Zoologische Jahrbücher, Abteilung für Systematik, Ökologie und Geographie der Tiere 109: 157 169. Borsuk-Bialynicka M, Moody SM. 1984. Priscagaminae, a new subfamily of the Agamidae (Sauria) from the Late Cretaceous of the Gobi Desert. Acta Palaeontologica Polonica 29: 51 81. Caujapé-Castells J, Jansen RK. 2003. The influence of the Miocene Mediterranean desiccation on the geographical expansion and genetic variation of Androcymbium gramineum (Cav.) McBride (Colchicaceae). Molecular Ecology 12: 1515 1525. Coppens Y. 1994. East side story: the origin of humankind. Scientific American 270: 62 69. Frost DR, Etheridge R. 1989. A phylogenetic analysis and taxonomy of iguanian lizards. Miscellaneous Publications of the University of Kansas Natural History Museum 81: 1 65. Fu J. 1998. Toward the phylogeny of the family Lacertidae: implications from mitochondrial DNA 12S and 16S gene sequences (Reptilia: Squamata). Molecular Phylogenetics and Evolution 9: 118 130. Geoffroy L, Huchon P, Khanbari K. 1998. Did Yemeni Tertiary granites intrude neck zones of a stretched continental upper crust? Terra Nova 10: 196 200.

258 S. A. M. AMER and Y. KUMAZAWA Girdler RW. 1991. The Afro-Arabian rift system. An overview. Tectonophysics 197: 139 153. Harzhauser M, Piller WE, Steininger FF. 2002. Circum- Mediterranean Oligo-Miocene biogeographic evolution the gastropods point of view. Palaeogeography, Palaeoclimatology, Palaeoecology 183: 103 133. Honda M, Ota H, Kobayashi M, Nabhitabhata J, Yong HS, Sengoku S, Hikida T. 2000. Phylogenetic relationships of the family Agamidae (Reptilia: Iguania) inferred from mitochondrial DNA sequences. Zoological Science 17: 527 537. Joger U. 1986. Phylogenetic analysis of Uromastyx lizards, based on albumin immunological distances. In: Rocek Z, ed. Studies in herpetology. Prague: Charles University, 187 191. Joger U. 1991. A molecular phylogeny of agamid lizards. Copeia 1991: 616 622. Joger U, Lambert MRK. 1996. Analysis of the herpetofauna of the Republic of Mali, 1. Annotated inventory, with description of a new Uromastyx (Sauria: Agamidae). Journal of African Zoology 110: 21 51. Kishino H, Hasegawa M. 1989. Evaluation of the maximum likelihood estimate of the evolutionary tree topologies from DNA sequence data, and the branching order in Hominoidea. Journal of Molecular Evolution 29: 170 179. Kocher TD, Thomas WK, Meyer A, Edwards SV, Pääbo S, Villablanca FX, Wilson AC. 1989. Dynamics of mitochondrial DNA evolution in animals: amplification and sequencing with conserved primers. Proceedings of the National Academy of Sciences, USA 86: 6196 6200. Kumazawa Y, Azuma Y, Nishida M. 2004. Tempo of mitochondrial gene evolution: Can mitochondrial DNA be used to date old divergences? Endocytobiosis and Cell Research 15: 136 142. Kumazawa Y, Endo H. 2004. Mitochondrial genome of the Komodo dragon: Efficient sequencing method with reptileoriented primers and novel gene rearrangements. DNA Research 11: 115 125. Kumazawa Y, Nishida M. 1993. Sequence evolution of mitochondrial trna genes and deep-branch animal phylogenetics. Journal of Molecular Evolution 37: 380 398. Kumazawa Y, Nishida M. 1995. Variations in mitochondrial trna gene organization of reptiles as phylogenetic markers. Molecular Biology and Evolution 12: 759 772. Kumazawa Y, Nishida M. 2000. Molecular phylogeny of osteoglossoids: a new model for Gondwanian origin and plate tectonic transportation of the Asian arowana. Molecular Biology and Evolution 17: 1869 1878. Kumazawa Y, Yamaguchi M, Nishida M. 1999. Mitochondrial molecular clocks and the origin of euteleostean biodiversity: Familial radiation of perciforms may have predated the Cretaceous/Tertiary boundary. In: Kato M, ed. The biology of biodiversity. Tokyo: Springer-Verlag, 35 52. MacDonald GM. 2003. Biogeography: introduction to space, time, and life. New York: John Wiley & Sons. Macey JR, Larson A, Ananjeva NB, Fang Z, Papenfuss TJ. 1997. Two novel gene orders and the role of light-strand replication in rearrangement of the vertebrate mitochondrial genome. Molecular Biology and Evolution 14: 91 104. Macey JR, Schulte JA II, Ananjeva NB, Larson A, Rastegar-Pouyani N, Shammakov SM, Papenfuss TJ. 1998. Phylogenetic relationships among agamid lizards of the Laudakia caucasia species group: testing hypotheses of biogeographic fragmentation and an area cladogram for the Iranian Plateau. Molecular Phylogenetics and Evolution 10: 118 131. Macey JR, Schulte JA II, Larson A, Ananjeva NB, Wang Y, Pethiyagoda R, Rastegar-Pouyani N, Papenfuss TJ. 2000. Evaluating trans-tethys migration: an example using acrodont lizard phylogenetics. Systematic Biology 49: 233 256. Mateo J, Geniez P, López-Jurado L, Bons J. 1998. Chronological analysis and morphological variations of saurians of the genus Uromastyx (Reptilia, Agamidae) in western Sahara. Description of two new taxa. Revista Española de Herpetologia 12: 97 109. McClanahan TR, Young TP. 1996. East African ecosystems and their conservation. Oxford: Oxford University Press. demenocal PB, Brown FH. 1999. Chapter 3: Pliocene tephra correlations between East African hominid localities, the Gulf of Aden, and the Arabian Sea. In: Agusti J, Rook L, Andrews P, eds. The evolution of Neogene terrestrial ecosystems in Europe. London: Cambridge University Press, 23 54. Mertens R. 1962. Bemerkungen über Uromastyx acanthinurus als Rassenkreis (Rept. Saur.). Senckenbergiana Biologica 43: 425 432. Moody SM. 1980. Phylogenetic and historical biogeographical relationships of the genera in family Agamidae (Reptilia: Lacertilia). PhD Thesis, University of Michigan, Ann Arbor, Michigan. Moody SM. 1987. A preliminary cladistic study of the lizard genus Uromastyx (Agamidae, sensu lato), with a checklist and diagnostic key to the species. In: van Gelder JJ, Strijbosch H, Bergers PJM, eds. Proceedings of the Fourth Ordinary General Meeting of the Society of European Herpetology, Nijmegen, Holland: Faculty of Sciences Nijmegen, 285 288. Posada D, Crandall KA. 1998. Modeltest: testing the model of DNA substitution. Bioinformatics 14: 817 818. Pudlo D, Shandelmeier H, Reynolds P-O. 1997. Chapter 16: The Late Tertiary (Langhian, ca. 15 Ma). In: Shandelmeier H, Reynolds P-O, eds. Palaeogeographicpalaeotectonic atlas of north-eastern Africa, Arabia, and adjacent areas. Late Neoproterozoic to Holocene. Rotterdam: A.A. Balkema, 105 110. Rögl F. 1998. Palaeogeographic considerations for Mediterranean and Paratethys seaways (Oligocene to Miocene). Annales Naturhistorisches Museum Wien 99A: 279 310. Ross DA, Schlee J. 1973. Shallow structure and geologic development of the southern Red Sea. Geological Society of America Bulletin 84: 3827 3843. Saitou N, Nei M. 1987. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular Biology and Evolution 4: 406 425. Schmidt HA, Strimmer K, Vingron M, von Haeseler A. 2002. TREE-PUZZLE: maximum likelihood phylogenetic analysis using quartets and parallel computing. Bioinformatics 18: 502 504.